The Tumor-suppressive role of CNTNAP2 in glioma: Dual regulation of the NDN-ERK axis and M2 macrophage polarization Authors: Yu Wu, Weiwei Zou, Shiqi Xia, Ti Zhang, Fan Yao, Yangqin Li, Xiaoqian Liu, Qian Peng, Xiao-Liang Xing Published: 2026 ABSTRACT AbstractCancer constitutes a persistent major global health burden, which underscores the urgent demand for innovative biomarkers and therapeutic targets. While the Contactin-associated protein (CNTNAP) family has been associated with tumor development, the pan-cancer role of CNTNAP2 is still poorly characterized. Researchers performed a comprehensive pan-cancer analysis of CNTNAP2 based on The Cancer Genome Atlas (TCGA) datasets to evaluate its expression profile and clinical relevance, and conducted functional experiments in U118MG glioblastoma cells and CNTNAP2-deficient mouse tissues to verify its biological function. The results identified CNTNAP2 as a dual diagnostic (AUC > 0.9) and prognostic biomarker for glioma, with low expression levels correlating to worse overall survival. In vitro assays showed that CNTNAP2 knockdown enhanced the proliferation and migration of U118MG cells, while its overexpression inhibited these malignant phenotypes. Mechanistically, CNTNAP2 deficiency triggered excessive activation of the ERK signaling pathway and downregulated its key downstream target Necdin (NDN), and pharmacological inhibition of ERK reversed the oncogenic effects caused by CNTNAP2 silencing. Further analysis demonstrated elevated M2 macrophage infiltration in glioma, and CNTNAP2 expression was closely correlated with M2 macrophage polarization; high CNTNAP2 expression also corresponded to lower TIDE scores, suggesting improved immunotherapy response. Collectively, CNTNAP2 acts as a tumor suppressor in glioma via the NDN-ERK axis, suppressing tumor progression and M2 macrophage polarization, making it a promising diagnostic, prognostic biomarker and potential therapeutic target for glioma.Subject terms: Biomarkers, Immunosuppression Introduction Cancer is a leading cause of human mortality and poses a major public health challenge worldwide. In 2022 alone, an estimated 20 million new cancer cases were diagnosed globally, resulting in 9.7 million deaths [1]. Approximately 53.5 million individuals were estimated to have survived within five years of their cancer diagnosis [1], underscoring the substantial burden of the disease. Notably, about one in five people will develop cancer during their lifetime [2], and projections indicate that by 2050, the global incidence of newly diagnosed cancers could reach 35 million per year [1, 3]. Therefore, the identification of biomarkers associated with cancer diagnosis, prognosis, and treatment is of profound clinical and practical importance. Cell adhesion molecules (CAMs), a class of cell-surface glycoproteins, play critical roles in diverse biological processes such as hemostasis, immune responses, inflammation, embryogenesis, and neural development [4–7]. Dysregulation of CAM function is closely implicated in tumor initiation, invasion, and metastasis [8–12]. The contactin-associated protein (CNTNAP) family, a subset of CAMs within the neurexin superfamily, primarily mediates neuronal and neuron–glial adhesion, which is essential for the development, structural maintenance, and function of the nervous system [12–14]. Emerging evidence, however, indicates that CNTNAPs are aberrantly expressed in various cancers. For example, CNTNAP1 is upregulated in kidney cancer [15], while CNTNAP2 shows abnormal methylation or is absent in glioma, myeloid leukemia, and pancreatic ductal adenocarcinoma [16–18]. Moreover, elevated CNTNAP5 expression is significantly associated with poorer overall survival in patients with renal cell carcinoma [19]. These observations collectively suggest important roles for CNTNAPs in oncogenesis. Nevertheless, research to date has been largely restricted to specific cancer types, and the pan-cancer functions of CNTNAPs remain largely unexplored. In this study, we conducted comprehensive bioinformatic analyses to evaluate CNTNAP family members as potential biomarkers across multiple cancer types and investigated their associations with the tumor immune microenvironment. Additionally, we employed in vitro and in vivo experimental approaches to elucidate the functional roles of CNTNAPs in cancer progression. Materials and methods The flow chart was shown in Supplementary Fig. 1. Public data acquisition and processing The original genetic data of various cancers and the corresponding clinical data were downloaded from the Cancer Genome Atlas (TCGA) database (downloaded on September 25, 2023). The validation data (GSE4290 and GSE68848) were downloaded from the Gene Expression Omnibus (GEO) database. For the analysis of differential expression and diagnostic value, only cancer types with at least three matched normal tissue samples were included. For the assessment of prognostic value, this study utilized 30 types of cancer. Gene expression data were normalized using the DESeq2 package in R 4.3.2. In its internal normalization algorithm, DESeq2 already incorporates correction for sample size. After that, differentially expressed genes (DEGs) were identified with the thresholds of |log2FoldChange | ≥ 0.5, Average Expression ≥ 0.5, and an adjusted p-value < 0.05. Subsequently, the DEGs were used for subsequent analyses using the normalized expression data by DESeq2, such as survival curves and correlation analyses. Animal studies CNTNAP2+/- mice (Stock #017482) were obtained from The Jackson Laboratory. All animal experiments were approved by the Ethics Committee of the Center for Medical Genetics, School of Life Sciences, Central South University, and were conducted in compliance with institutional guidelines and the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Genotyping was performed using the primers in Table 1. Table 1The primers sequence information.CategorySequence (5’ to 3’)Wild-type forwardCTGCCAGCCCAGAACTGGWild-type reverseCGCTTCCTCGTGCTTTACGGTATMutant reverseACACCAGGGGCAAGAATTG The primers sequence information. Survival and enrichment analysis Patients were categorized into high- and low-expression groups based on the median expression level of each CNTNAPs gene. Using the packages “survival”, “survminer”, and “RegParallel” in the R 4.3.2 platform, univariate and multivariate Cox regression analysis was conducted to screen the overall survival rate-related signatures for CNTNAPs. We performed univariate Cox regression analysis to identify CNTNANPs genes potentially associated with overall survival (OS) in patients, followed by multivariate Cox regression analysis to determine which of these could serve as independent prognostic factors. Functional enrichment analysis of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways was performed using the DAVID bioinformatics resource (v6.8) with default parameters [20]. The gene set used for enrichment analysis was specifically the list of significantly DEGs identified from differentially expressed analysis by DESeq2. For significantly enriched pathways, we applied DAVID’s Bonferroni-corrected p-value with a threshold of p < 0.05 as the criterion for significance. Cell culture and transfection U118MG glioblastoma cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin at 37°C in a 5% CO₂ atmosphere. For overexpression, cells were transfected with a pcDNA3.1 vector encoding CNTNAP2 (Zeqiong Biotech, China). For the knockout experiment, the researchers transfected the specific small interfering RNA (siRNA) targeting CNTNAP2 into the cells according to the instructions in the manuscript (Universal Biotech, China). The pharmacological inhibitor PD98059 was purchased from MedChemExpress (USA). For co-culture experiments, THP-1 monocytes were differentiated into M0 macrophages by treatment with 100 ng/mL phorbol 12-myristate 13-acetate (PMA) for 48 h in the lower chamber of a 12-well plate. After washing with phosphate-buffered saline (PBS), fresh medium was added. U118MG cells, pre-transfected with either si-CNTNAP2 or a negative control siRNA (si-NC), were trypsinized, resuspended in serum-free medium (2×10⁵ cells/insert), and seeded into Transwell inserts placed above the differentiated M0 macrophages. The co-culture was maintained for 48 h. Subsequently, macrophages were harvested for RNA extraction, and the mRNA expression levels of M1 (CD80, TNF-α, iNOS) and M2 (CD163, CD206, Arg-1) polarization markers were analyzed by Quantitative real-time PCR. RNA sequencing Total RNA was isolated from the hippocampal tissues of wild-type (n = 3, age 6–8 weeks) and CNTNAP2-/- (n = 3, age 6–8 weeks) mice using TRIzol™ reagent (Thermo Fisher Scientific). The RNA-seq analysis was conducted by NovaGene (China) Company following the Illumina standard protocol and was sequenced. In each processing step, the samples must undergo quality control verification before moving on to the next analysis. Quantitative real-time PCR MRNA was extracted using the RNA Easy Fast Kit (TIANGEN, China) according to the manufacturer’s instructions. Two μg of total RNA was reverse transcribed using Thermo Scientific RevertAid™Kit (Thermo Scientific, USA). Quantitative real-time PCR (qRT-PCR) was performed using the Light Cycler®480 Instrument II (Roche). The gene expression was normalized with control as the standard, and then a differential expression analysis was conducted. The primers were showed as follow (Table 2). Table 2The primers sequence information.GeneForward Primer (5’ to 3’)Reverse Primer (5’ to 3’)CD80ACCCCCAACATAACTGAGTCTTTCCAACCAAGAGAAGCGAGGTNF-αCTCTTCTGCCTGCTGCACTTTGATGGGCTACAGGCTTGTCACTCiNOSGCTCTACACCTCCAATGTGACCCTGCCGAGATTTGAGCCTCATGCD163CCAGAAGGAACTTGTAGCCACAGCAGGCACCAAGCGTTTTGAGCTCD206GTTTCCATCGAGACTGCTGCGCCACTTTCCTTCAACATTTCGArg-1TCATCTGGGTGGATGCTCACACGAGAATCCTGGCACATCGGGAAGAPDHGTCTCCTCTGACTTCAACAGCGACCACCCTGTTGCTGTAGCCAA The primers sequence information. Western blot analysis Tissues or cells were lysed in 2× SDS loading buffer (50 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol) containing protease inhibitors (NaF, Na3VO4, PMSF). Proteins were separated by SDS-PAGE, transferred to PVDF membranes, and blocked with 5% non-fat milk in TBST (Tris-buffered saline with 0.1% Tween-20). Membranes were incubated with primary antibodies overnight at 4°C, followed by incubation with appropriate HRP-conjugated secondary antibodies. Protein bands were visualized using an enhanced chemiluminescence (ECL) detection system. The following primary antibodies were used: anti-phospho-ERK (T202/Y204) (Abways, CY5277), anti-ERK (Abways, CY5277), anti-CNTNAP2 (Abcam, ab33994), anti-Necdin (Abcam, ab18554), and anti-GAPDH (Thermo Fisher, ET1601-4). The protein expression was normalized with control as the standard, and then a differential expression analysis was conducted. Functional assays Cell Viability: Cell viability was assessed using a CCK-8 assay kit (Beyotime Biotechnology, Shanghai, China) at 24-, 48-, and 72-h post-transfection. 10 μL of CCK-8 reagent added to 100 μL of medium per well. Absorbance was measured at 450 nm, 2 h after adding the CCK-8 reagent. Wound Healing Assay: Confluent cell monolayers (90%) were scratched with a sterile pipette tip. After washing with PBS, cells were cultured in serum-free medium. The migration distance was measured and quantified at 0 and 24 h. we used an inverted microscope (OLYMPUS, CKX53) to take images of the scratch area at specified magnifications (40) at 0 h and 24 h. The width of the scratch was quantified using the ImageJ software. Colony Formation Assay: In the 6-well plate, 1000 cells were seeded per well based on manual counting. The transfected cells were plated in 6-well plates and cultured for 14 days. Afterward, the cells were fixed with 100% methanol, stained with 0.5% crystal violet, and the number of colonies (defined as >50 cells) was manually counted. Colony formation was visualized using an inverted microscope (OLYMPUS, CKX53) and analyzed with ImageJ software. Statistical analysis Data are presented as mean ± standard error of the mean (SEM). Statistical comparisons between two groups were performed using an unpaired, two-tailed Student’s t-test. For comparisons involving more than two groups, one-way or two-way ANOVA with Bonferroni’s post-hoc test was applied. A p-value of less than 0.05 was considered statistically significant. Results Diagnostic and prognostic value of CNTNAP family members in pan-cancer Analysis of the Human Protein Atlas database revealed distinct, tissue-specific expression patterns across the CNTNAP family (Fig. 1a). CNTNAP1, CNTNAP2, and CNTNAP3 exhibited relatively high expression levels, whereas CNTNAP4 and CNTNAP5 were expressed at lower levels (Fig. 1a). Diagnostic value of the CNTNAP family across human cancers. a Expression profiles of CNTNAP family members in paracancerous human tissues, based on the Human Protein Atlas database. b Heatmap showing the expression levels of CNTNAPs in tumor tissues and adjacent normal tissues for different cancer types. BLCA (Paracancerous = 19, Cancer = 408), BRCA (Paracancerous = 113, Cancer = 1098), CESC (Paracancerous = 3, Cancer = 306), CHOL (Paracancerous = 9, Cancer = 36), COAD (Paracancerous = 41, Cancer = 458), ESCA (Paracancerous = 11, Cancer = 162), Glioma (Paracancerous = 5, Cancer = 692), HNSC (Paracancerous = 44, Cancer = 502), KICH (Paracancerous = 24, Cancer = 65), KIRC (Paracancerous = 72, Cancer = 531), KIRP (Paracancerous = 32, Cancer = 289), LIHC (Paracancerous = 50, Cancer = 373), LUAD (Paracancerous = 59, Cancer = 515), LUSC (Paracancerous = 49, Cancer = 501), PCPG (Paracancerous = 3, Cancer = 183), PRAD (Paracancerous = 52, Cancer = 496), READ (Paracancerous = 10, Cancer = 167), STAD (Paracancerous = 32, Cancer = 375), THCA (Paracancerous = 58, Cancer = 510), UCEC (Paracancerous = 35, Cancer = 544). c Heatmap showing the expression levels of CNTNAPs in alive and dead cancer patients. d The expression profiles of CNTNAP2 and CNTNAP4 in glioma patients and the control group in TCGA database. e The expression profiles of CNTNAP2 and CNTNAP4 in alive and dead glioma patients. ROC curve of CNTNAP2 f and CNTNAP4 g in glioma diagnosis. h The expression profiles of CNTNAP2 in glioma patients and the control group in GSE4290 database. i ROC curve of CNTNAP2 in glioma diagnosis in GSE4290 database. j The expression profiles of CNTNAP2 in glioma patients and the control group in GSE68848 database. k ROC curve of CNTNAP2 in glioma diagnosis in GSE68848 database. *p < 0.05, **p < 0.01, ***p < 0.001. Following exclusion of cancer types with fewer than three adjacent normal samples, differential expression analysis revealed significant dysregulation of CNTNAP genes in multiple cancers, with a predominant pattern of downregulation (Fig. 1b). Similarly, CNTNAP genes were significantly dysregulated in tissues from deceased cancer patients, also showing a trend toward decreased expression (Fig. 1c). Comprehensive analysis identified CNTNAP2 and CNTNAP4 as consistently expressed in glioma, where both genes exhibited significant downregulation. These genes demonstrated considerable diagnostic potential, with area under the curve (AUC) values of 0.952 for CNTNAP2 and 0.875 for CNTNAP4. Furthermore, their expression was significantly reduced in glioma tissues from deceased patients. Kaplan-Meier survival analysis indicated that glioma patients with low expression of CNTNAP2 had significantly worse overall survival (OS), disease-specific survival (DSS), disease-free interval (DFI), and progression-free interval (PFI) (Supplementary Fig. 2a–2d, left). Similarly, low expression of CNTNAP4 was associated with poorer OS, DSS, and PFI, though no significant difference was observed for DFI (Supplementary Fig. 2a–2d, right). The prognostic characteristics of CNTNAP2 and CNTNAP4 are further summarized in Supplementary Fig. 2e–2h. The expression of CNTNAP2 was significantly correlated with the OS, DSS, DFI, and PFI of glioma. Genetic alteration landscape of CNTNAP genes We characterized genetic alterations in CNTNAP genes using cBioPortal and the TCGA/ICGC Pan-Cancer Analysis of Whole Genomes dataset. CNTNAP2 displayed the highest alteration frequency (9%), primarily through amplifications, deep deletions, and missense mutation (Supplementary Fig. 3a). CNTNAP showed different mutation patterns in various types of cancers. Among them, CNTNAP1 had the highest alteration frequency in embryonic tumors, pancreatic cancer, and endometrial cancer, and its main alteration type is amplification (Supplementary Fig. 3b). CNTNAP2 had the highest alteration frequency in melanoma, non-small cell lung cancer, and lung cancer. The alterations of CNTNAP2 in melanoma were mainly mutations and amplifications. The alterations of CNTNAP2 in non-small cell lung cancer were mainly mutations, amplifications, and deep deletions. The alterations in lung cancer were mainly mutations and amplifications (Supplementary Fig. 3c). CNTNAP3 had the highest alteration frequency in bladder cancer, lung cancer, and pancreatic cancer. The alterations of CNTNAP3 in bladder cancer were mainly amplifications. The alterations of CNTNAP3 in lung cancer and pancreatic cancer were mainly amplifications and deep deletions (Supplementary Fig. 3d). CNTNAP4 had the highest alteration frequency in bladder cancer, melanoma, and non-small cell lung cancer. The alterations of CNTNAP4 in bladder cancer were mainly amplifications. In melanoma, the mainly alteration was mutations. In non-small cell lung cancer, the mainly alteration were mainly mutations and amplifications (Supplementary Fig. 3e). The alteration frequency of CNTNAP5 was the highest in non-small cell lung cancer, melanoma and colorectal cancer. The main type of alteration was mutation (Supplementary Fig. 3f). CNTNAP2 modulates glioma progression via the ERK signaling pathway To investigate the functional role of CNTNAP2 in glioma, we performed knockdown and overexpression experiments in U118MG cells (Fig. 2a). CNTNAP2 knockdown significantly enhanced cell viability (Fig. 2b), migration (Fig. 2d, e), and colony formation (Fig. 2g, h). Conversely, CNTNAP2 overexpression suppressed these malignant phenotypes (Fig. 2c, f, and i). CNTNAP2 suppresses malignant phenotypes in Glioma cells. a Western blot analysis confirming the efficiency of CNTNAP2 knockdown and overexpression in U118MG cells. b, c Cell viability assessed by CCK-8 assay after CNTNAP2 knockdown b and overexpression c. Data are presented as mean ± SEM. n = 3. Cell migration evaluated by wound healing assay. Schematic diagram d and quantitative results of migration after CNTNAP2 knockdown e and overexpression f. n = 3. Cell proliferation assessed by colony formation assay. Schematic diagram g and quantitative results of colony formation after CNTNAP2 knockdown h and overexpression i. n = 3. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001. Mechanistically, we identified 6641 differentially expressed genes (DEGs) between CNTNAP2-high and CNTNAP2-low glioma patients (Fig. 3a). Pathway enrichment analysis of these DEGs revealed 38 significantly altered pathways (Bonferroni p < 0.05), with the top 10 shown in Fig. 3b. Correlation analysis identified 12,011 genes significantly associated with CNTNAP2 expression; the 10 strongest positive and negative correlates are displayed in Fig. 3c. Enrichment analysis of these correlated genes identified 145 pathways, with 18 showing significant enrichment (top 10 in Fig. 3d). Nine pathways were common to both analyses, including the MAPK signaling pathway (hsa04010). CNTNAP2 suppresses Glioma progression via ERK pathway inhibition. a Volcano plot of differentially expressed genes (DEGs) between CNTNAP2‑high and CNTNAP2‑low glioma samples. b Top 10 KEGG pathways enriched in DEGs from a. c Top 10 genes most positively and negatively correlated with CNTNAP2 expression in glioma. d Top 10 KEGG pathways enriched among CNTNAP2‑correlated genes from c. e Schematic of the ERK signaling pathway. f, g Western blot f and quantification g of p‑ERK and t‑ERK in U118MG cells after CNTNAP2 modulation. The protein expression was normalized with control as the standard. n = 3. h Schematic of ERK signaling in vivo. i, j Western blot i and quantification j of p‑ERK and t‑ERK in hippocampal tissues from WT vs. CNTNAP2-/- mice. The protein expression was normalized with control as the standard. n = 6. k Schematic of PD98059‑mediated ERK inhibition. l, m Western blot l and quantification m showing PD98059 reverses p‑ERK elevation induced by CNTNAP2 knockdown. The protein expression was normalized with control as the standard. n = 3. n CCK‑8 assay: PD98059 rescues CNTNAP2 knockdown‑induced proliferation. n = 3. o, p Wound‑healing schematic o and quantitation p: PD98059 blocks CNTNAP2 knockdown‑enhanced migration. n = 3. q, r Colony formation assay schematic q and quantitation r: PD98059 suppresses CNTNAP2 knockdown‑driven proliferation. n = 3. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001. Given the oncogenic relevance of MAPK signaling, we assessed ERK phosphorylation (p-ERK) in U118MG cells following CNTNAP2 modulation. CNTNAP2 knockdown markedly increased p-ERK levels, while its overexpression reduced them (Fig. 3e–g). Consistent with prior reports of increased astrocyte proliferation in CNTNAP2-/- mice, we found significantly elevated hippocampal p-ERK levels in these animals (Fig. 3h–j). Pharmacological inhibition of ERK phosphorylation using PD98059 attenuated the pro-tumorigenic effects induced by CNTNAP2 knockdown: it suppressed p-ERK activation (Fig. 3k–m) and reversed the enhancements in proliferation, migration (Fig. 3n–r). CNTNAP2 autonomously regulates the expression of NDN RNA-seq analysis of CNTNAP2-/- mouse hippocampi identified 583 DEGs (320 upregulated, 263 downregulated; Fig. 4a). Through a cross-analysis of DEGs from CNTNAP2-/- mouse models and glioma patients with low CNTNAP2 expression, we identified 13 genes that are consistently dysregulated in both cohorts (Fig. 4b, c) and showed significant correlation with CNTNAP2 expression (Fig. 4d, 4e). Univariate Cox analysis indicated that all 13 genes were associated with overall survival in glioma (Fig. 4f), with CD80, EMILIN3, ESCO2, ESM1, GDNF, NDN, RBP3, SLC25A43, and TRPM8 demonstrating strong prognostic value (AUC > 0.7; Fig. 4g). CNTNAP2 autonomously regulates NDN expression. a Volcano plot of DEGs in hippocampal tissues of CNTNAP2-/- vs. WT mice. b, c, Heatmaps of overlapping gene expression in glioma patients (b) and in WT vs. CNTNAP2-/- mice c. d, e Correlation bubble plots of CNTNAP2 with overlapping genes in glioma d and mouse hippocampus e. f Forest plot of univariate Cox regression for overlapping genes in glioma overall survival. g ROC curves of overlapping genes for prognostic prediction in glioma. h Schematic of CNTNAP2 modulation and NDN measurement. i Western blot and quantification: CNTNAP2 knockdown decreases and overexpression increases NDN protein in U118MG cells. The protein expression was normalized with control as the standard. n = 3. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001. Necdin (NDN) was initially identified as a hypothetical protein, which was found in mouse embryonic cancer cells. It is expressed only through the epigenetic imprinting mechanism from the paternal chromosome. NDN, as an intracellular inhibitory factor, can inhibit the activity of the EGFR/ERK signaling pathway [21]. Additionally, previous study indicated that the intracellular segment of CNTNAP2 promotes the nuclear translocation of calcium/calmodulin-dependent serine protein kinase, thereby regulating the transcription of genes such as NDN [22]. Based on previous studies, we examined whether the activation of the ERK signaling pathway caused by CNTNAP2 is related to NDN, and confirmed this regulatory relationship in U118MG cells: Knockdown of CNTNAP2 leads to a significant downregulation of NDN expression, while overexpression of CNTNAP2 results in a significant increase in NDN expression (Fig. 4h, i). CNTNAP2 knockdown promotes M2 macrophage polarization Given the association of CNTNAP2 with immune-related pathways, we evaluated its impact on the glioma immune microenvironment. We analyzed the immune status of glioma using different algorithms. The infiltration of immune cells is shown in Fig. 5. We found that in all three algorithms, the immune infiltration of macrophage M2 was significantly increased in glioma patients (Fig. 5a–c). The results of the correlation analysis showed that CNTNAP2 was significantly negatively correlated with Macrophage M2 (Fig. 5d–f). The WGCNA analysis revealed that in the TCGA dataset, the 6 modules based on the LM22, ConsensusTME and Cibersort algorithms were associated with macrophage M2, with 6 modules and 2 modules respectively showing such an association (Supplementary Fig. 5a–5c). Based on the parameters |MM | ≥ 0.3 and |GS | ≥ 0.3, 1368 genes are associated with macrophage M2 (Supplementary Fig. 5j). In the GSE4290 dataset, based on the LM22, ConsensusTME and Cibersort algorithms, there are 6 modules. Among them, 6 modules and 7 modules are respectively associated with macrophage M2 (Supplementary Fig. 5d–5f). Based on the parameters |MM | ≥ 0.3 and |GS | ≥ 0.3, 3,939 genes are associated with macrophage M2 (Supplementary Fig. 5k). In the GSE68848 dataset, based on the LM22, ConsensusTME and Cibersort algorithms, there are 6 modules. Each of the 6 modules is respectively associated with macrophage M2 (Supplementary Fig. 5g–5i). Based on the parameters |MM | ≥ 0.3 and |GS | ≥ 0.3, 8,244 genes are associated with macrophage M2 (Supplementary Fig. 5l). A comprehensive analysis revealed that 828 genes are associated with Macrophage M2, including CNTNAP2 (Supplementary Fig. 5m). CNTNAP2 expression correlates with immune landscape in Glioma. a–c Macrophage M2 immune infiltration of glioma patients from the TCGA a, GSE4290 b, and GSE68848 c dataset based on different algorithms. Correlation of CNTNAP2 with macrophage M2 of glioma patients from the TCGA d, GSE4290 e, and GSE68848 f dataset based on different algorithms. mRNA expression of M1 g and M2 h markers in macrophages co‑cultured with si‑NC or si‑CNTNAP2 U118MG cells. The protein expression was normalized with control as the standard. n = 3. i K-M curve of macrophage M2 in OS. Left, TCGA. Middle, GSE4290. Right, GSE68848. j K-M curve of macrophage M2 in DSS. Left, TCGA. Middle, GSE4290. Right, GSE68848. k K-M curve of macrophage M2 in DFI. Left, TCGA. Middle, GSE4290. Right, GSE68848. l K-M curve of macrophage M2 in PFI. Left, TCGA. Middle, GSE4290. Right, GSE68848. m TIDE, Dysfunction, and Exclusion scores compared between immunotherapy response groups. n Correlation of CNTNAP2 expression with TIDE and Exclusion scores. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001. In vitro co-culture experiments validated this finding: CNTNAP2 knockdown in U118MG cells co-cultured with THP-1-derived M0 macrophages significantly decreased the expression of M1 markers (CD80, TNFA, iNOS; Fig. 5g) while increasing the expression of M2 markers (CD163, CD206, ARG1; Fig. 5h). The Kaplan-Meier survival analysis indicated that in glioma patients, high M2-type macrophage infiltration was significantly associated with overall survival rate (OS), disease-specific survival rate (DSS), and progression-free interval (PFI). These prognostic associations were further verified in the TCGA, GSE4290, and GSE68848 datasets (Fig. 5i, j, and l). DFI was associated with high M2-type macrophage infiltration in the GSE4290 dataset (Fig. 5k). Tumor Immune Dysfunction and Exclusion (TIDE) analysis revealed that CNTNAP2-high glioma patients exhibited significantly lower TIDE, Dysfunction, and Exclusion scores (Fig. 5m), suggesting a greater potential for response to immunotherapy. CNTNAP2 expression showed significant negative correlations with TIDE and Exclusion scores (Fig. 5n). Discussion Cell adhesion molecules (CAMs), a class of cell-surface proteins, mediate interactions between cells and the extracellular matrix (ECM) and are critical for maintaining tissue architecture [5]. Dysregulation of CAM function is closely associated with tumor initiation, invasion, and metastasis [10, 11]. The CNTNAP family, members of the neurexin superfamily of CAMs, are known to play vital roles in neuronal development. Recent evidence suggests that CNTNAPs are also implicated in cancer [15–19]. In this study, we performed a pan-cancer analysis to systematically investigate the expression and functional significance of CNTNAP family members. Analysis of the Human Protein Atlas database revealed distinct expression patterns of CNTNAPs across human tissues. CNTNAP1, CNTNAP2, and CNTNAP3 were widely expressed at relatively high levels, whereas CNTNAP4 and CNTNAP5 showed generally low expression, with notable enrichment in the nervous and urinary systems—consistent with previous reports [23]. This pattern suggests important biological stratification within the family. Based on existing literature, we hypothesize that CNTNAP1, CNTNAP2, and CNTNAP3 play dominant roles in constructing basic neural circuits and myelin formation, whereas CNTNAP4 and CNTNAP5 are more specialized in regulating specific neuronal types and neurotransmitter circuits. The high expression of CNTNAP1, CNTNAP2, and CNTNAP3 likely reflects their fundamental roles in brain development. CNTNAP1 is anchored in the lateral zone of the axon hillock, which is crucial for maintaining structural integrity and enabling rapid action potential conduction [24]. CNTNAP2 interacts with CNTN2 and potassium channels, localizing to sub-regions below the axonal initial segment, where it regulates neuronal excitability and network synchrony; its dysfunction is linked to neurological disorders such as autism and epilepsy [25, 26]. CNTNAP3 has been implicated in inhibitory synapse maturation [27]. Thus, the high expression of these three proteins aligns with the developmental demand for myelination and basic circuit assembly. In contrast, CNTNAP4 and CNTNAP5 exhibit low but specific expression; for instance, CNTNAP4 is present in dopaminergic neurons and modulates dopamine release via presynaptic mechanisms [28], indicating functional specialization within specific neurotransmitter systems. We analyzed the mutation status of CNTNAP genes and found that CNTNAP2 exhibited a relatively high alteration frequency. Mutations in CNTNAP2 can disrupt multiple functions, contributing to a spectrum of disorders from neurodevelopmental conditions to malignant tumors [16, 29–32]. In the nervous system, CNTNAP2 is essential for neuronal migration, dendritic development, and excitatory/inhibitory balance [25, 26, 33], with mutations or reduced expression representing significant risk factors for autism and epilepsy [29, 34, 35]. In cancer biology, CNTNAP2 functions as a tumor suppressor [16, 30]. Its large genomic structure renders it susceptible to inactivation via deletion, mutation, or methylation in various cancers, and its low expression is consistently associated with poor patient prognosis [30]. Thus, investigating CNTNAP2 reveals shared molecular pathways between neurological disorders and cancer and provides a foundation for diagnostic and therapeutic strategies. Differential expression analysis indicated that CNTNAP genes are frequently downregulated across multiple tumor types, though cancer-specific upregulation was also observed. For instance, CNTNAP1 was significantly elevated in KIRC and KIRP, aligning with prior findings [15]. CNTNAP2 was downregulated in glioma tissues and further decreased in deceased patients, corroborating earlier reports of its aberrant expression in glioma [16]. Additionally, CNTNAP2 knockout mice exhibit not only autism-like behaviors but also hippocampal astrocyte proliferation [26], suggesting a role in cell growth regulation. Using TCGA data, we evaluated the prognostic significance of CNTNAPs. High CNTNAP1 expression was associated with poor survival in renal cancer [15], and CNTNAP5 was linked to overall survival in the same cancer type [19]. Previous studies reported focal rearrangements and reduced CNTNAP2 expression in recurrent oligodendrogliomas, with expression levels inversely correlating with tumor grade [30]. Our results strengthen the evidence that low CNTNAP2 expression is closely linked to glioma progression. Loss of CNTNAP2 function is strongly linked to neurodevelopmental disorders, and emerging evidence suggests its pathological effects extend to glioma pathogenesis through shared pathways involving neural circuit dysfunction and glial responses. CNTNAP2 deficiency leads to early neuronal migration abnormalities and reduced interneuron numbers [26], disrupting excitatory/inhibitory (E/I) balance and causing abnormal network activity in regions such as the hippocampus. Clinically, this manifests as epilepsy, learning deficits, and repetitive behaviors [25, 26]. This cascade of neural circuit abnormalities overlaps with the clinical presentation of hippocampal glioma, where patients also exhibit memory decline, attentional deficits, and epilepsy [36, 37]. This suggests that the hyperexcitable and unstable neural network resulting from CNTNAP2 deficiency creates a permissive microenvironment for glioma development. Chronic neuronal hyperexcitability and epileptic activity persistently activate astrocytes, leading to reactive astrogliosis [38]. Activated astrocytes release inflammatory cytokines and growth factors, creating a state of chronic inflammation [38, 39]. This microenvironment exacerbates neuronal damage and circuit abnormalities and may promote glial cell proliferation and malignant transformation, thereby contributing to glioma initiation and progression [39, 40]. To directly assess the functional impact of CNTNAP2 loss, we performed in vitro assays in U118MG glioblastoma cells. Knockdown of CNTNAP2 significantly enhanced cell proliferation and migration, whereas its overexpression suppressed these malignant behaviors, indicating a tumor-suppressive role in glioma, consistent with phenotypes observed in CNTNAP2-/- mice [26]. Mechanistically, enrichment analysis of CNTNAP2-associated genes and differentially expressed genes in CNTNAP2-low tumors indicated involvement of the MAPK signaling pathway. RNA sequencing of hippocampal tissue from CNTNAP2-/- mice identified NDN as a downstream target, positively correlated with CNTNAP2 expression. This regulatory relationship was confirmed in U118MG cells, where CNTNAP2 knockdown reduced and overexpression increased NDN levels. These findings suggest that CNTNAP2 positively regulates NDN expression, consistent with reports that the CNTNAP2 intracellular domain (CICD), generated by γ-secretase cleavage, interacts with CASK and translocates to the nucleus to regulate NDN transcription [22]. However, whether this represents direct regulation requires further investigation. NDN, a paternally expressed imprinted gene, acts as a tumor suppressor in several cancers, including urothelial carcinoma, melanoma, breast cancer, and colorectal cancer [41–44]. The reduction of NDN following CNTNAP2 loss, coupled with glial proliferation, implies that NDN may exert anti-cancer effects in glioma. Fujimoto et al. reported that in cortical progenitor cells lacking NDN, astrocyte differentiation induced by cardiotrophin-1 was accelerated in the presence of EGF, an effect abolished by inhibition of EGFR/ERK signaling [21]. We therefore hypothesized that CNTNAP2 deficiency promotes astrocyte activation via the NDN-ERK pathway. Consistent with this, CNTNAP2 knockdown led to ERK pathway activation, while its overexpression suppressed ERK activation. In vivo, hippocampal tissue from CNTNAP2-/- mice showed significant ERK activation. Treatment with an ERK inhibitor markedly reduced glial cell proliferation and migration, confirming that CNTNAP2 regulates these processes via the ERK pathway [21, 22]. As a member of the immunomodulatory superfamily, CNTNAP2 is also associated with immune modulation, particularly M2 macrophage polarization. Previous studies indicate that ERK activation promotes M2 polarization [45, 46], which fosters tumor progression [47, 48]. In glioma samples, low CNTNAP2 expression was associated with increased M2 macrophage infiltration and poorer survival. In vitro co-culture experiments showed that CNTNAP2 knockdown in U118MG cells skewed macrophages toward an M2 phenotype, as indicated by altered expression of M1 and M2 markers. Thus, CNTNAP2 may influence tumor immunity by modulating ERK-mediated M2 polarization, though precise mechanisms require further exploration [49, 50]. Notably, CNTNAP1 and CNTNAP2 belong to the same protein family. CNTNAP1 expression correlates positively with infiltration of cancer-associated fibroblasts, regulatory T cells, and myeloid-derived suppressor cells in renal cancer, and is significantly associated with M2 macrophages in renal clear cell carcinoma [15]. This supports the plausibility of CNTNAP2’s role in M2 polarization. We also found a significant negative association between CNTNAP2 expression and the TIDE score, a computational framework modeling tumor immune evasion and predicting response to immune checkpoint blockade (ICB) [51]. Specifically, CNTNAP2 correlated negatively with the exclusion score, reflecting impaired cytotoxic T lymphocyte infiltration into tumors. Low CNTNAP2 expression may thus contribute to an immune-excluded phenotype, associated with poor ICB response across multiple cancers [51, 52]. This study has several limitations. First, mechanistic and functional validations were primarily conducted in the U118MG cell line; future studies should include a broader panel of glioma cell lines and patient-derived primary cells to confirm generalizability. Second, conclusions regarding the role of CNTNAP2 role in the tumor immune microenvironment rely largely on bioinformatic predictions and in vitro co-culture assays, lacking direct validation in in vivo tumor models. To address this, we plan to establish orthotopic glioma models to directly investigate how CNTNAP2 modulates tumor growth and the immune landscape in vivo. This constitutes a key focus for our subsequent research. Conclusion Our study establishes CNTNAP2 as a pivotal tumor suppressor and prognostic biomarker in glioma. We demonstrate that its downregulation predicts poor patient outcomes and promotes tumor malignancy by disrupting the NDN-ERK axis. Specifically, CNTNAP2 sustains NDN expression to constrain ERK signaling, and its loss leads to hyperactive ERK-driven tumor progression, a phenotype reversible upon ERK inhibition. Furthermore, CNTNAP2 shapes an anti-tumor immune microenvironment by restraining M2 macrophage polarization. These findings unveil a dual mechanism of CNTNAP2 action, encompassing direct pathway suppression and immune modulation, thereby underscoring its significant therapeutic relevance. Supplementary information Supplementary Information Author contributions Xiao-Liang Xing conceived and designed the experiments; Yu Wu, Weiwei Zou, Shiqi Xia, and Ti Zhang performed the analysis and experiment; Fan Yao, Yangqin Li, Xiaoqian Liu, and Qian Peng helped to perform the analysis; Xiao-Liang Xing wrote the paper. Yu Wu, Weiwei Zou, Shiqi Xia, Ti Zhang contributes equally to this work. Funding We acknowledge the Hunan Provincial Department of Education (22A0710), The Science and Technology Program of Hunan Province (2021JJ40389, 2022JJ40294, 2023JJ50442), College student innovation and entrepreneurship project (S202412214009) for funding the research. Data availability The datasets generated and analyzed for the current study are available. Competing interests The authors declare no competing interests. Ethics approval and consent to participate All methods were performed in accordance with approved guidelines. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Yu Wu, Weiwei Zou, Shiqi Xia, Ti Zhang. Supplementary information The online version contains supplementary material available at 10.1038/s41417-026-01046-3. References Associated Data Supplementary Materials Supplementary Information Data Availability Statement The datasets generated and analyzed for the current study are available.